Impact-induced hydrothermal activity on early Mars

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 110,, doi: /2005je002453, 2005 Impact-induced hydrothermal activity on early Mars Oleg Abramov and David A. Kring Lunar and Planetary Laboratory, University of Arizona, Tucson, Arizona, USA Received 15 April 2005; revised 5 August 2005; accepted 15 August 2005; published 4 November [1] We report on numerical modeling results of postimpact cooling of craters with diameters of 30, 100, and 180 km in an early Martian environment, with and without the presence of water. The effects of several variables, such as ground permeability and the presence of a crater lake, were tested. Host rock permeability is the main factor affecting fluid circulation and lifetimes of hydrothermal systems, and several permeability cases were examined for each crater. The absence of a crater lake decreases the amount of circulating water and increases the system lifetime; however, it does not dramatically change the character of the system as long as the ground remains saturated. It was noted that vertical heat transport by water increases the temperature of localized near-surface regions and can prolong system lifetime, which is defined by maximum near-surface temperature. However, for very high permeabilities this effect is negated by the overall rapid cooling of the system. System lifetimes, which are defined by near-surface temperatures and averaged for all permeability cases examined, were 67,000 years for the 30-km crater, 290,000 years for the 100-km crater, and 380,000 for the 180-km crater. Also, an approximation of the thermal evolution of a Hellas-sized basin suggests potential for hydrothermal activity for 10 Myr after the impact. These lifetimes provide ample time for colonization of impact-induced hydrothermal systems by thermophilic organisms, provided they existed on early Mars. The habitable volume reaches a maximum of 6,000 km 3 8,500 years after the impact in the 180-km crater model. Citation: Abramov, O., and D. A. Kring (2005), Impact-induced hydrothermal activity on early Mars, J. Geophys. Res., 110,, doi: /2005je Introduction 1.1. Impact-Induced Hydrothermal Systems [2] Impact-induced hydrothermal systems are initiated by an impact event into a water-rich or ice-rich target material. A fraction of the kinetic energy of the impactor raises the temperature of the planetary crust, providing a thermal driver for the circulation of water and the release of steam and other volatiles such as CO 2. In particular, several longterm heat sources are generated during an impact event: shock-heated solid material, impact melt, and the central uplift. In every hypervelocity impact, a shock wave is generated in the target, compressing the material and depositing a large amount of internal energy. Since a shock is not thermodynamically reversible, waste heat is produced upon decompression of the shocked material by the rarefaction waves that follow [e.g., Ahrens and O Keefe, 1972; Kieffer and Simonds, 1980]. A phase change can occur if enough heat is deposited, resulting in the melting or vaporization of the target rocks. If sufficient melt is produced, a melt sheet is formed in the crater basin. Additionally, material below the central region of the crater is uplifted during the formation process, delivering warm material closer to the surface. The relative importance of these heat sources depends on the diameter of the crater. The dominant heat source for small, simple craters (less than 7 km in diameter on Mars) Copyright 2005 by the American Geophysical Union /05/2005JE002453$09.00 is shock-emplaced heat, due to negligible amounts of melt and uplift. For larger, complex craters, melt sheets and central uplifts become important, with melt sheets contributing significantly more energy than central uplifts [Daubar and Kring, 2001; Thorsos et al., 2001]. [3] Evidence of impact-induced hydrothermal activity is present at several terrestrial impact craters in the form of mineral assemblages indicative of high-temperature hydrothermal alteration. Examples of known alteration sites include, in order of increasing diameter, the 4-km Kärdla crater [Versh et al., 2003], the 24-km Haughton crater [Osinski et al., 2001], the 35-km Manson crater [e.g., McCarville and Crossey, 1996], the 80-km Puchezh- Katunki crater [e.g., Naumov, 1993, 2002], the 180-km Chicxulub crater [Kring and Boynton, 1992; Ames et al., 2004; Hecht et al., 2004; Zürcher and Kring, 2004] and the 150- to 250-km Sudbury crater [e.g., Farrow and Watkinson, 1992; Ames et al., 1998]. Impact-induced hydrothermal activity has been suggested for Martian craters as well [Newsom, 1980; Allen et al., 1982; Newsom et al., 1996]. [4] While there are no known active impact-induced hydrothermal systems today, their presence may have been dramatically greater at 3.9 Ga. Several lines of evidence indicate that the inner solar system was subjected to a sharp increase in the number of impacts at that time. Analysis of lunar crust samples [e.g., Turner et al., 1973; Tera et al., 1974] and impact melts [e.g., Dalrymple and Ryder, 1993, 1996; Cohen et al., 2000] returned by the Apollo and Luna 1of19

2 missions, as well as meteorites, indicates that the rocks in the lunar crust were either thermally metamorphosed or melted at 3.9 Ga. This is interpreted as a consequence of a dramatic increase in the number of impacts in a relatively brief time span of 20 to 200 Myr [e.g., Tera et al., 1974; Ryder, 2000]. This cataclysm was not limited to the Earth- Moon system as meteorites from multiple bodies in the asteroid belt, as well as the only sample of the ancient Martian crust (meteorite ALH 84001), show effects of impact-induced metamorphism at 3.9 Ga [Bogard, 1995; Kring and Cohen, 2002] Importance of Impact-Induced Hydrothermal Systems on Early Mars [5] The sites of ancient hydrothermal systems are very promising places to search for biomarkers indicative of past biological activity on Mars. Of the two types of hydrothermal systems, volcanogenic and impact-induced, the latter may have been more important early in Mars s history, because old (Noachian-age) terrains are dominated by impact cratering processes and show little evidence of volcanism [Carr, 1996]. This implies that impact cratering removed volcanic landforms that may have been produced during this period of high heat flow, and was therefore the dominant process for some time. Also, while volcanic process on Mars have produced positive topographic features, such as the Tharsis shield volcanoes, impact events formed basins that could have filled with water to form crater lakes [e.g., Newsom et al., 1996; Cabrol et al., 2001], enhancing their biological potential. Finally, during the impact cataclysm at 3.9 Ga, the heat delivered by impact events may have exceeded that generated by volcanic activity [Kring, 2000]. [6] Ancient valley networks, some of which are interpreted as surface runoff [e.g., Williams and Phillips, 2001; Craddock and Howard, 2002], as well as chemical, mineralogical, and structural data from the Opportunity rover [Squyres et al., 2004], indicate that liquid water was present and stable on the surface (at least episodically) in the Noachian epoch, during which the cataclysm took place. The cataclysm may have resurfaced Mars, forming most of the craters observed in the Martian highlands [Kring and Cohen, 2002], and likely would have resulted in cycles of vaporization of any surface water or ice following very large impacts (forming craters at least 600 km in diameter) [Segura et al., 2002], eliminating any life that may have existed at the surface. At the same time, new subsurface habitats were created [Zahnle and Sleep, 1997] in the form of impact-induced hydrothermal systems, which may have provided sanctuary to existing life or were a site of its origin. There are several lines of evidence suggesting that may have been the case on Earth. For example, phylogenies of terrestrial organisms constructed from rrna sequences imply a thermophilic or hyperthermophilic common ancestor [e.g., Pace, 1997], which, along with the earliest isotopic evidence of life at 3.85 Ga [Mojzsis and Harrison, 2000] coinciding with the cataclysm, suggest that impact-induced hydrothermal systems played an important role in the origin and evolution of early life on Earth. The same may be true for Mars. It is also important to note, however, that impactgenerated hydrothermal systems were not limited to early Mars. Present-day subsurface ice has been inferred at high latitudes (poleward of 60 ) on the basis of the detection of hydrogen by the Gamma Ray Spectrometer (GRS) on board Mars Odyssey [Boynton et al., 2002], and indirectly by the presence of fresh craters with fluidized ejecta blankets [e.g., Mouginis-Mark, 1987; Squyres et al., 1992; Barlow and Perez, 2003] and rootless cones [Lanagan et al., 2001] at lower latitudes. Thus a present-day impact may still generate hydrothermal activity. [7] Unlike Earth, many ancient Martian craters are wellpreserved and presumably contain a mineralogical record of hydrothermal activity. Future remote sensing missions, such as the upcoming Mars Reconnaissance Orbiter (MRO), will have spectrometers with high spatial resolution, potentially capable of detecting localized outcrops of hydrothermally altered lithologies. In anticipation of these missions, it is vital to model the spatial distribution of the alteration in Martian craters, predict mineral assemblages indicative if hydrothermal activity, and outline a search strategy. 2. Goals [8] One of the goals of this work is to constrain the lifetimes of impact-induced hydrothermal systems on early Mars. The duration of hydrothermal activity can affect the magnitude of chemical and mineralogical alteration of the Martian crust and is a strong factor in determining the potential biological importance of these systems. As discussed qualitatively by Newsom et al. [2001], the system lifetime depends on the size and initial temperature of the heat source, physical and thermal parameters of host rocks (such as permeability, heat capacity, and thermal conductivity), availability of liquid water, and other factors, such as self-sealing due to mineral precipitation. Crater cooling models suggest that the lifetimes of hydrothermal systems in craters 20 to 200 km in diameter are 10 3 to 10 6 years if purely conductive cooling is assumed [e.g., Daubar and Kring, 2001; Turtle et al., 2003; Ivanov, 2004]. In order to better constrain the expected lifetimes of these systems we are using a finite difference computer simulation to evaluate the additional effects of heat transport by water and steam. [9] Another goal is to further understand the mechanics of postimpact hydrothermal circulation, with a focus on the locations of near-surface activity. This in turn can aid in spectroscopic and visual identification of hydrothermal vents and hydrothermally altered minerals at Martian craters. [10] Finally, we are seeking to understand the biological potential of these systems in terms of their habitable volume, or the rock volume within which there are temperatures and fluid flow suitable for thermophilic microorganisms. The evolution of habitable volume as a function of time can place important constraints on the long-term habitability of these systems. [11] The present work aims to achieve these goals by numerically modeling hydrothermal activity in craters of 30, 100, and 180 km diameter in an early Martian environment. In addition, a conductive cooling timescale for a Hellas-size impact basin is estimated. 3. Model 3.1. Computer Code HYDROTHERM [12] A modified version of the publicly available program HYDROTHERM (source code available from 2of19

3 authors) is used for the simulations in this work. HYDROTHERM is a three-dimensional finite difference computer code developed by the U.S. Geological Survey to simulate water and heat transport in a porous medium [Hayba and Ingebritsen, 1994]. Its operating range is 0 to 1200 C and 0.5 to 1,000 bars; however, the upper temperature limit has been extended by the authors to 1700 C for the modeling of impact melt sheets. For a more detailed theoretical discussion of the thermodynamics and fluid mechanics of hydrothermal modeling and their implementation in HYDROTHERM, please refer to Abramov and Kring [2004] and references therein. [13] HYDROTHERM has been previously used for scientific applications; in particular, it was applied to hydrothermal systems of volcanic origin on Earth [e.g., Hayba and Ingebritsen, 1997] and Mars [Gulick, 2001], as well as hydrothermal systems at terrestrial [Abramov and Kring, 2004] and Martian [Rathbun and Squyres, 2002] impact craters. This work presents several specific improvements on Rathbun and Squyres [2002]. While the earlier simulations were limited to 50, ,000 years, in this work hydrothermal activity is simulated for up to several million years, allowing estimates of system lifetimes for larger craters. Crater lakes and the latent heat of fusion are now explicitly included in the model. The crater topography has been improved on the basis of observations of lunar craters, and is preserved throughout the simulation, rather than being removed shortly after crater formation. In addition, the postimpact temperature distributions that serve as starting conditions have been generated by hydrocode simulations either specifically for Mars or were adapted for Mars with consideration of the different kinetic energy requirements for the formation of Martian craters. [14] In general, HYDROTHERM is well suited for modeling impact-induced hydrothermal systems, although the code relies on several assumptions. Perhaps its most significant shortcoming for the purposes of this work is its inability to model brines. The program makes the assumption that the fluid is pure water, while hydrothermal fluids generally contain some dissolved solids. In particular, several studies point to the existence and composition of Martian brines. These include laboratory simulations of brine formation based on the mineralogy of the SNC meteorites [Bullock et al., 2004] and a compositional analysis of the Nakhla meteorite [Sawyer et al., 2000], which suggests that it has been in contact with a seawaterlike brine or a hydrothermal fluid. System parameters, like the boiling point, depend on the concentration of these solutes, which in turn depends on a variety of factors such as the amount of water available, surface and subsurface temperatures, erosion rates, atmospheric pressure and composition, etc., which are not well quantified for early Mars. If we assume the bulk composition of water on early Mars is similar to terrestrial seawater, then its thermodynamic properties at subcritical temperatures are sufficiently close to that of pure water for the purposes of this model. At supercritical temperatures, the effect of solutes in H 2 Ois minimized by extremely low permeabilities at those temperatures [Hayba and Ingebritsen, 1997]. Another important assumption made by HYDROTHERM is that the rock and water are in a local thermal equilibrium. This assumption is valid if the fluid flow is relatively slow and steady, and breaks down in cases of rapid transients. For this breakdown to occur, water would have to pass through 300 to 500 meters of rock (vertical resolution of our models) without reaching equilibrium. Such transients are unlikely except perhaps in the very early stages of the system. HYDROTHERM also assumes that the ground remains fully saturated throughout the simulation, meaning that all pore spaces remain filled by water or steam. With the exception of the elevated rim, which drains rapidly after crater formation, this is a fine assumption because the surface of a crater lake represents the water table and the ground below that datum is expected to be fully saturated. Meanwhile the permeability and porosity of the crater rim are set to near-zero to simulate an unsaturated elevated surface. Water fluxes through the rim due to atmospheric precipitation are highly uncertain and are not included in the model, but, for an annual precipitation of less than 10 cm, would be significantly lower than the hydrothermal fluxes Model Conditions [15] Taking advantage of an impact crater s radial symmetry, we examine a vertical cross section from the center of the crater to beyond the outer rim. All models are represented on a grid, with a total of 2,475 blocks. The upper boundary of the model represents a layer of cooled breccia, with pressure and temperature held constant at 0.5 bars and 1 C. The thickness of the breccia layer is equal to the vertical resolution of the model, which is 333 m for the 30-km crater and 500 m for the 100- and 180-km craters. It is effectively an infinite source or sink of the fluid, donating or accepting water depending on underlying hydrologic conditions. It also functions as a heat sink; so when the thermal energy reaches the upper boundary, it is permanently removed from the system. This construct is reasonable in a rapidly convecting crater lake situation, where heat is rapidly removed from the upper surface layer, and water is freely exchanged. The bottom boundary is impermeable with a constant basal heat flux of 32.2 mw/m 2 to match an average geothermal gradient of 13 C km 1 [Babeyko and Zharkov, 2000]. The lefthand boundary of the model is the axis of symmetry and is thus impermeable and insulating. The right-hand boundary is permeable for both fluid and heat and is located sufficiently far away from the center of the crater that the temperatures are close to an average geothermal gradient. The depth of the models varies, with 10 km for the 30-km crater, 15 km for the 100-km crater, and 14.5 km for the 180 km crater, and it was found by trial and error that further extending the depth had no appreciable effect on hydrothermal activity. This can also be shown analytically using an expression for the thermal conductive time: t ¼ z2 rc p ; ð1þ k where z is depth and r, C p and k are density, heat capacity, and thermal conductivity, respectively, given in Table 2. This expression indicates that the time required for the heat to conductively propagate to the surface from a depth 3of19

4 Figure 1. (a) A 29-km fresh Martian crater located at 23 N, 207 E, just west of Tharsis (Viking MDIM). (b) An elevation profile of the crater (running north to south), which was used to provide topography for the 30-km crater model. Derived from a global MOLA digital elevation model gridded at 128 pixels per degree. of 10 km is close to 3 million years, well over the duration of hydrothermal activity. 4. Input Parameters 4.1. Topography The 30-km Crater [16] The topography for the 30-km model crater was obtained directly from the Mars Orbiter Laser Altimeter (MOLA) data for a 29-km fresh Martian crater located at 23 N, 207 E (Figure 1). This crater has been characterized as very young by Mouginis-Mark et al. [2003], on the basis of the presence of ejecta rays, extensive secondary craters, high thermal inertia, lack of superimposed small impact craters, large depth/diameter ratio, and other geomorphic indicators. In addition, the rim height of this crater is 680 m and the central peak height is 820 m, which is significantly higher than 340 m and 220 m, respectively, predicted by Garvin et al. [2003] on the basis of topography analysis of a large sample of Martian craters. In fact, topography derived from lunar craters (Table 1), which predicts the rim and central peak heights of 920 m and 490 m, respectively, provides a better fit. This is likely an indication of an advanced degree of degradation of most Martian craters The 100-km and 180-km Craters [17] While the recent MOLA data on the topography of large Martian craters is of very high quality, it is likely not representative of the state of the craters immediately after their formation, since the analyzed craters been degraded in various degrees by erosional and depositional processes. It is also noteworthy that the topography of the 29-km fresh crater is better approximated by lunar morphometry than by MOLA-derived predictions. Therefore our topography models for the 100- and 180-km craters are based on the morphometry of lunar craters, summarized in Table Temperature Distribution The 30-km Crater [18] The temperature distribution for the 30-km crater model was generated by a hydrocode simulation of crater formation on Mars [Pierazzo et al., 2004]. The simulation modeled a 90 impact of an asteroid 2 km in diameter at a velocity of 8 km/s. A hydrocode simulation for a cometary impact, which produced higher temperatures, was also available, but was not used because asteroid impacts are far more prevalent. Since the geothermal gradient chosen for our models (13 C/km) matched the geothermal gradient in the Pierazzo et al. [2004] simulation, no further changes to the temperature distribution were made The 100-km and 180-km Craters [19] The temperature distribution for the 100-km crater model was obtained from a hydrocode simulation of the Table 1. Parameters Used for Reconstruction of the Original Topography for Craters With D = 30, D = 100, and D = 180 km a Parameter Dependence on Rim-to-Rim Diameter (D, km) Source Crater depth D Pike [1977] Crater floor diameter 0.19 D 1.25 Pike [1977] Peak ring diameter 0.5 D Wood and Head [1976] Peak ring height 3 Hale and Grieve [1982] Peak ring thickness 0.11 D Pike [1985] Rim height (0.236 D )((0.5D) 3 /r 3 ) Pike [1977], Melosh [1989] a The variable r is the distance from the center of the crater. All parameters were obtained from morphometric studies of lunar craters. After Melosh [1989]. 4of19

5 Figure 2. Final crater diameter as a function of projectile kinetic energy for Mars and Earth. Calculated from the Pi-group scaling laws, assuming a stony impactor with a density of 3,000 kg/m 3, a target density of 2,600 kg/m 3, and a competent rock or saturated soil target type. Impact velocities typical for asteroid impacts are used: 7 km/s for Mars and 17 km/s for Earth. impact that formed Popigai crater on Earth [Ivanov, 2004], and adapted for Mars. An important factor that was considered was that the kinetic energy of the projectile needed to form a 100-km crater on Mars is lower compared to Earth (Figure 2). The main reasons for this are the lower gravity on Mars, which allows more material to be ejected, and the lower average impact velocity for Mars, which requires a larger projectile to generate the same kinetic energy if the density stays constant. Figure 2 indicates that the kinetic energy needed to form a crater of a given diameter is 50% less for Mars than for Earth. Since the amount of shock heating and uplift is approximately proportional to the kinetic energy of the projectile, this implies that immediately after an impact a 100-km crater on Mars would be 50% cooler than a 100-km crater on Earth. Consequently, the net temperature increase (DT) in the Popigai crater temperature distribution was reduced by 50%. [20] Similarly, the temperature distribution for our 180-km crater model was obtained from a hydrocode simulation of Sudbury crater formation [Ivanov and Deutsch, 1999] and adapted for early Mars using the methods outlined in the previous paragraph Hellas Basin (2,000 km) [21] Hydrocode simulations for the formation of the Hellas Basin have not yet been conducted. Consequently, the temperature distribution for a Hellas-sized impact basin (2,000 km in diameter) was computed analytically using an expression for specific waste heat (DE w ) derived from the Murnaghan equation of state by Kieffer and Simonds [1980]: DE w ¼ 1 2 PV 0 2K " 0V 0 1 Pn # 1=n þ 1 þ K 0V 0 " n K 0 nð1 nþ 1 Pn # 1 ð1=nþ þ 1 ; ð2þ K 0 where P is the peak shock pressure, K 0 is the adiabatic bulk modulus at zero pressure, n is the pressure derivative of the bulk modulus, and V 0 is the specific uncompressed volume (1/r 0 ). For basalt, the uncompressed density r 0 is 2600 kg/m 3, K 0 is 19.3 GPa, and n is 5.5 [Gault and Heitowit, 1963]. Shock pressure P drops off with distance r from the impact point according to the power law P ¼ A r n ; ð3þ where R pr is the radius of the projectile, n is the decay exponent, and A is pressure at r = R pr [e.g., Pierazzo and Melosh, 2000]. R pr was estimated at 91,500 m using Pi-group scaling laws, assuming a vertical asteroid impact with a velocity typical for Mars (7,000 m/s) into competent rock. R pr 5of19

6 For a 90 impact, the decay exponent n was estimated at ± by Pierazzo and Melosh [2000] using hydrocode simulations. A can be estimated by A ¼ r 0 ½C þ Su 0 Šu 0 ; ð4þ where C and S are constants, with C = 2600 m/s and S = 1.62 [Melosh, 1989, p. 232]. Assuming that the density of the projectile is roughly equivalent to that of the target, the initial particle velocity u 0 can be calculated by rffiffiffiffiffiffiffi 1 u 0 ¼ ; ð5þ 2 v2 where v is the impactor velocity [Melosh, 1989, p. 65], in this case 7,000 m/s. To obtain the final temperature increase, specific waste heat DE w is divided by the heat capacity, which is 800 J kg 1 K 1 for basalt [Mellon, 2001]. [22] The temperature at the base of the lithosphere is estimated at 1300 C[Schubert et al., 2001], which gives a lithospheric thickness of 100 km with a 13 C/km temperature gradient. The temperature gradient in the mantle is taken to be 0.1 C/km [Schubert et al., 2001]. The final temperature distribution obtained using this method agrees well with a hydrocode-generated temperature distribution for a hypothetical impact of a 200-km body on the Earth [Ivanov, 2004] Melt Sheet Properties [23] The volume of melt sheets in the 100- and 180-km craters was estimated using an analytical expression derived by S. M. Wong et al. (Differential melt scaling for oblique impacts on the Earth, Moon, and Mars, submitted to Meteoritics and Planetary Science, 2005): V melt ¼ 1: r p r t g 0:18 D 0:83 tc D 2:35 pr v 1:63 sin 1:63 q; ð6þ where r p is the density of the projectile (3,000 kg/m 3 ), r t is the density of the target (2600 kg/m 3 ), g is the acceleration due to gravity (3.72 m/s 2 ), D tc is the transient crater diameter as measured at the pre-impact surface, D pr is the projectile diameter, v is the impact velocity (7,000 m/s), and q is the impact angle (90 ). [24] Several expressions for relating the final crater diameter and the transient crater diameter have been put forward. Croft [1985] suggested the following relationship based on the terrace widths and central peak diameters of lunar craters: D ¼ D1:18 tr DQ 0:18 ; ð7þ where D is the rim-to-rim diameter of the final crater, D tr is the rim-to-rim diameter of the transient crater, and D Q is the simple-to-complex transition diameter. D Q is estimated to be 8,400 m for early Mars assuming inverse scaling with gravity and density [Holsapple, 1993]: D Q ¼ g Moonr Moon D QMoon gr t : ð8þ Kring [1995] derived a different expression for the transient/ final crater relation based on the morphology of ejecta blankets (in SI units): D ¼ 0:82D 1:07 tr : ð9þ Croft [1985] and Kring [1995] expressions eventually intersect at large diameters but strongly disagree at small diameters (D tr D Q ) at which the relationship should be close to the simple crater relation D = 1.19D tr [Melosh, 1989]. Thus, for small complex craters, the Croft [1985] expression underestimates the final crater diameter D while the Kring [1995] expression overestimates it. For that reason, we use a geometric mean of the two expressions: D ¼ 0:91 D1:125 tr D 0:09 ; ð10þ Q 1.13 which also agrees with the proportionality D / D tr proposed by McKinnon and Schenk [1985], who estimated the degree of collapse on the basis of the ratio of ejecta blanket diameter to the crater diameter. Equation (9) yields transient diameters of 62,000 m and 105,000 m for the 100-km crater, and the 180-km crater, respectively. [25] However, equation (5) requires the transient crater diameter measured at the original surface (D tc, also called the apparent diameter), while equation (9) yields the rim-torim diameter D tr of the transient crater. Pike [1977] determined the ratio of the rim-to-rim diameter to the apparent diameter for 164 lunar craters, finding that it decreases with increasing crater diameter, ranging from 1.20 for craters less than 0.4 km to 1.11 for craters greater than 100 km. Since there is little collapse in very small craters, we assume that the ratio D tr /D tc is close to 1.20, which agrees with hydrocode simulations of transient crater formation by Shuvalov et al. [2002]. It is also close to the average of the ratios suggested by Holsapple [1993] (1.30) and Grieve and Garvin [1984] (1.13). The apparent transient craters diameters are then 52,000 m and 88,000 m for the 100- and 180-km craters, respectively. [26] The projectile diameter D pr is calculated using the Pigroup scaling laws, resulting in diameters of 8,500 m for the 100-km crater and 16,500 m for the 180-km crater. This then yields total melt volumes of 460 km 3 and 3,400 km 3 for the 100- and 180-km craters, respectively. However, a substantial fraction of the melt is ejected from the crater. For lunar craters in the 100 to 180 diameter range, the fraction of melt ejected is estimated at 0.5 [Cintala and Grieve, 1998]. The fraction of melt ejected scales with gravity and is estimated at 0.24 for the terrestrial crater Chicxulub [Kring, 1995]. However, this value is not well defined and strongly depends on the assumed value of z in the z-model of Maxwell [1977]. Therefore, using a conservative approach, we assume a lunar value of 1/2 melt ejected. The remaining melt volumes of 230 km 3 and 1700 km 3 are distributed in central melt sheets and small melt sheets, the latter of which are located between the peak rings and rims in the annular trough of the craters. [27] The initial temperature of the melt was conservatively estimated at 1700 C on the basis of temperature estimates in excess of 1700 C for melt sheets of terrestrial impact craters 6of19

7 Table 2. Rock Parameters Used in the Model Parameter Value Units Porosity f(z), 20% at the surface unitless Permeability f(z,t), 10 2 at the surface a darcies Thermal conductivity (crust) 2.5 W m 1 K 1 Heat capacity (crust) 800 J kg 1 K 1 Density (crust) 2600 kg m 3 Thermal conductivity (mantle) 3.3 W m 1 K 1 Heat capacity (mantle) 1250 J kg 1 K 1 Density (mantle) 3300 kg m 3 a Unless otherwise indicated. [e.g., Grieve et al., 1977; Ostermann et al., 1996]. The melt/clast ratio in the melt sheet increases with crater diameter and is expected to be high for the 180-km crater. However, for the 100-km crater, the volume proportion of cold clasts becomes an important factor in determining the initial volume and temperature of the melt sheet. The proportion of clasts in the melt sheet of a similarly sized Popigai crater on Earth is 50% [Masaitis et al., 1998]. Thus the corresponding volume of clasts was added to the melt sheet of the 100-km Martian crater, doubling its volume to 460 km 3 and lowering its temperature to 1125 C after thermal equilibration. [28] The latent heat of fusion is included in the model using the approximation of Jaeger [1968], after Onorato et al. [1978], replacing the heat capacity C p in the temperature range between the liquidus (T L ) and the solidus (T S ) with Cp 0 ¼ C p þ L= ðt L T S Þ: ð11þ [29] Here, L is the latent heat of fusion of diopside, 431 kj kg 1. The liquidus and solidus temperatures of 1280 C and 1070 C, respectively, were chosen on the basis of those of gabbro [Ernst, 1976], a coarse-grained equivalent of basalt, because the surface of Mars appears to be dominated by basaltic lithologies [Bandfield et al., 2000]. [30] An important question pertaining to the melt sheets is how much convection occurs during the cooling process. There is no heat source below the melt sheet, since the underlying rocks have a cooler temperature and act as a heat sink. Thus cooling occurs at both the top and bottom of the melt sheet, and the former may generate some convective overturn in the form of cold downwelling plumes below the stagnant lid. This mechanism has been observed at the Makaopuhi lava lake in Hawaii but for various reasons is not active at other lava lakes [Davaille and Jaupart, 1993]. Convection is not expected to occur in the melt sheet of the 100-km crater, since its initial temperature is well below the liquidus, a temperature at which convection in top-cooled magma reservoirs typically ceases [e.g., Brandeis and Marsh, 1989]. While some convection may have occurred in the melt sheet of the 180-km crater, potentially cooling it faster, it was not included in the model due to its uncertain extent and duration and the limitations of the code Rock Parameters [31] The porosity in our model decreases exponentially with depth, accounting for the closing of pore spaces by lithostatic pressure, following the approach suggested by Binder and Lange [1980] for the lunar crust: FðÞ¼F z 0 expð z=kþ; ð12þ where F 0 is surface porosity (20%) and K is the decay constant, which scales with gravity and is 1.07 km for Earth and 2.80 km for Mars [Clifford, 1993]. The depth z is measured with respect to local topography, not the preimpact surface level. [32] It is reasonable to assume that the number of fractures in an impact crater decreases with depth [e.g., Nordyke, 1964], and thus permeability in our model decays exponentially with depth similarly to porosity. Due to lower gravity on Mars, the depth to the base of the fractures is 2.5 times what it would be on Earth and permeability decreases more gradually with depth. Permeability is also a function of temperature, approximating the effect of the brittle/ductile transition at about 360 C [Fournier, 1991] by log linearly decreasing permeability with increasing temperature between 360 and 500 C: kz ðþ¼k 0 expð z=kþ T < 360 C log kz ðþþ11 log kz; ð TÞ ¼ ð500 TÞ T 500 C k ¼ darcies T > 500 C: ð13þ [33] Since permeability has important effects on the dynamics and duration of a hydrothermal system, and is not well constrained, several values of k 0 were investigated in this study. It was also recognized that lower permeabilities may become more appropriate due to mineralization as the system ages. Rock density, thermal conductivity, and heat capacity were assumed to be that of basalt [Mellon, 2001] and are summarized in Table 2. Additionally, the physical properties of the mantle for the Hellas basin simulation were taken from Schubert et al. [2001] Crater Lakes [34] Lakes may have commonly formed in impact craters on early Mars: one survey, conducted by Cabrol and Grin [1999] suggested paleolakes occurred in at least 179 craters. Whether these crater lakes can form shortly after crater formation on a timescale significantly less than the lifetime of the hydrothermal system depends on hydrologic conditions and the amount of atmospheric precipitation. A crater lake would certainly form rapidly if the impact occurred into an area that had abundant surface water, such as that inferred for an ancestral stage of the Meridiani Planum region on the basis of observations by MER Opportunity [Squyres et al., 2004]. A crater lake can also form through groundwater drainage from an underground aquifer, as observed at Meteor Crater on Earth [Shoemaker and Kieffer, 1974] and suggested for the 150-km Gusev crater on Mars [Grin and Cabrol, 1997]. In addition, Newsom et al. [1996] suggested that crater lakes may form in large craters (>65 km in diameter) shortly after an impact and persist even under current climatic conditions. Consequently, crater lakes may be a common phenomena, particularly on early Mars, and were incorporated into the models presented in this paper. 7of19

8 Figure 3. Results of a numerical simulation of the hydrothermal system at a 30-km crater on early Mars. The central peak of the crater is on the left side of each figure. Surface permeability k 0 is 10 2 darcies. Solid arrows and dotted arrows indicate the water and steam fluxes, respectively. The lack of arrows in some regions indicates that fluxes are less than 2 orders of magnitude smaller than the maximum flux. Solid lines are isotherms, labeled in degrees Celsius. The length of the arrows scales logarithmically with the flux magnitude, and the maximum value changes with each plot. (a) 25 years, max. water flux = kg s 1 m 2, max. steam flux = kg s 1 m 2 ; (b) 1,000 years, max. water flux = kg s 1 m 2, max. steam flux = kg s 1 m 2 ; (c) 10,000 years, max. water flux = kg s 1 m 2 ; (d) 100,000 years, max. water flux = kg s 1 m 2. However, a model was also run without a crater lake to define its role in hydrothermal activity. 5. Results 5.1. Hydrothermal System Mechanics and Lifetimes The 30-km Crater [35] Results of the numerical simulation of a hydrothermal system in a 30-km crater are shown in Figure 3. Overall, this system is driven by a central hot spot, which heats up water and causes it to rise, generating a single large upwelling that persists throughout the system s lifetime. At 25 years the system is primarily characterized by a large region within the crater s central peak where temperatures and pressures are compatible with water s gaseous phase. This results in emission of large quantities of steam up to three kilometers (horizontally) from the center of the crater. Water is drawn toward the crater s center to replenish the escaping steam. While not modeled explicitly due to the limitations of the code, this phase transition would have left behind minerals and caused a degree of clogging in this region. However, subsequent flow of hot water through the central peak may have redissolved these minerals. The source and sink of the water is the crater lake; virtually no water is drawn from the permeable right boundary after the formation of the crater lake. [36] By 1,000 years, the steam emission from the central peak has essentially ceased. While there are still small quantities of steam being generated within the central peak, it condenses before reaching the surface. The temperatures 8of19

9 Figure 4. Results of a numerical simulation of the hydrothermal system at a 100-km crater on early Mars. The center of the crater is on the left side of each figure. Surface permeability k 0 is 10 2 darcies. Solid arrows and dotted arrows indicate the water and steam fluxes, respectively. The lack of arrows indicates that fluxes are less than 2 orders of magnitude smaller than the maximum flux. Solid lines are isotherms, labeled in degrees Celsius. The length of the arrows scales logarithmically with the flux magnitude, and the maximum value changes with each plot. (a) 500 years, max. water flux = kg s 1 m 2, max. steam flux = kg s 1 m 2 ; (b) 4,000 years, max. water flux = kg s 1 m 2, max. steam flux = kg s 1 m 2 ; (c) 20,000 years, max. water flux = kg s 1 m 2, max. steam flux = kg s 1 m 2 ; (d) 200,000 years, max. water flux = kg s 1 m 2. within the central peak are now noticeably cooler, having decreased by more than 100 C. Due to this smaller thermal driver, the water fluxes are about 5 times smaller than they were at 25 years. However, the system on the whole still remains quite hot, with temperatures of over 100 C observed in the near-surface up to 6 km away from the crater s center. [37] At 10,000 years, steam generation has completely stopped, and temperatures and water fluxes have further decreased. The center of the single giant convection cell that makes up the system has started moving downward, a trend that continues in subsequent time steps. This allows water to recirculate without entering the lake. Finally, at 100,000 years, a small amount of circulation continues, but the water fluxes are 2 orders of magnitude lower than they were early in the system The 100-km Crater [38] Evolution of an impact-induced hydrothermal system in a 100-km crater is shown in Figure 4. Unlike the 30-km crater model, this crater is large enough to possess a high-temperature and impermeable melt sheet and temperatures high enough to render parts of the subsurface impermeable. One similarity to the 30-km crater, which had steam emission from the central peak, is that the steam is being generated and emitted from the peak ring early in the system, and is being replenished by water flowing in on both sides of the peak ring (not shown). However, this activity ceases by 500 years due to 9of19

10 the rapid cooling of the peak ring, and is replaced by a strong upwelling of water within the peak ring, where cold water enters near the base and is subsequently heated and transported upward. The highest fluxes of the 500-year time step are observed in this region. At this time in the simulation, most of the melt in the annular trough has crystallized, although it is not yet permeable to fluid flow. Some steam is being generated in the crater s modification zone on its outer wall, and there are some strong but somewhat chaotic water fluxes in this area as well, that are probably venting through the faults that separate blocks of crust in the modification zone. The flow of water through these faults likely reorganized into conduits by the precipitation of silica, phyllosilicates, and other minerals and subsequent self-sealing, a phenomena observed at mid-ocean ridge black smokers [e.g., Sleep, 1991] and most subaerial hydrothermal systems. If the temperature drops below the freezing point, formation of ice can have the same effect. Unlike the 30-km crater model, a small fraction of the circulating water is supplied by the permeable right boundary and not the crater lake. However, these fluxes about 2 orders of magnitude smaller than those seen elsewhere in the system. [39] At 4,000 years, the small melt sheet has completely crystallized and cooled. Several convection cells have developed in the annular trough and the modification zone, and there are several obvious deflections of the temperature contours due to the upward flow of warmer water and downward flow of colder water. The central melt sheet has fully crystallized, but remains completely impermeable due to high temperatures. The strong upwelling in the peak ring continues to be very active. Also, relatively large steam fluxes are seen within the peak ring. Surprisingly, the overall flux magnitudes are now higher than they were at 500 years, probably due to a smaller impermeable rock volume that was previously impeding flow. [40] At 20,000 years, most of the central melt sheet has cooled below 360 C and is now permeable to water. There is now vigorous activity near the right edge of the former central melt sheet with a couple of prominent convection cells. The upwelling in the peak ring continues as before, and a small upward flow in the modification zone on the far wall of the crater becomes noticeable as well. By 200,000 years, the character of the system has changed. The convection cells at the site of the central melt sheet have merged into two strong upwellings, which are the dominant feature of this time step. Far smaller upwellings continue to trickle inside the peak ring and along the far wall. However, they are essentially insignificant, and the overall character of the system is now close to that of the 30-km crater: a strong central upward flow driven by a central hot spot. The central upwelling continues until the end of the system when temperatures become too low to sustain it The 180-km Crater [41] Figure 5 shows the evolution of a hydrothermal system in 180-km crater. The main structural features here are similar to those of the 100-km crater, with larger central and annular melt sheets, higher topography, and higher temperatures in the central region. At 4,000 the central melt sheet has reached its liquidus temperature and is undergoing crystallization, while the small melt sheet has fully solidified. The peak ring again plays host to a prominent upwelling that continues throughout the lifetime of the system. Another long-lived but relatively weak upwelling develops at the outer wall in the crater s modification zone, and is being resupplied by water both from the lake and the permeable right boundary. There is some steam generation deep within the crater below the right edge of the central melt sheet, with steam originating near the critical point of water at 374 C. [42] At 20,000 years, the small melt sheet has completely crystallized, but a remnant hot spot remains and drives an upwelling in this area. The central melt sheet has completely crystallized and is partly permeable, allowing water to start circulating above it. There is also some supercritical steam being produced below the central melt sheet. [43] At 200,000 years, the central melt sheet has cooled and is fully permeable, allowing several large convection cells to develop in this region. This is a major difference from the 100-km crater model, where these convection cells at this time step were replaced by a single central upwelling. The convection cells continue to operate until the end of the system. The upwellings within the peak ring and the outer wall continue as before, but with reduced fluxes. The overall magnitude of the fluxes continues to decrease, with the largest flux observed here being 2 times smaller than those seen early in the system. [44] At 2 million years the system has long ceased operating and temperatures have returned close to a geothermal gradient. Only a hint of higher temperatures remains in the center of the crater. There are a couple of fossil flows still active, most notably in the modification zone, but the fluxes are 20 times smaller than those seen earlier in the system. This weak circulation, driven mainly by surface relief, may persist for a long time. It should also be noted, however, that these fluxes may be even smaller or entirely nonexistent due to fracture closing by hydrothermal mineralization and clay deposition, processes that generally operate on timescales smaller than those required for the cooling of this large crater. Given the low volumes of circulating water and the volume of rocks they must traverse, water would be expected to cool off completely before reaching the surface Effects of Crater Lake [45] For the purposes of comparison, a simulation was run without a crater lake present in the crater basin, but with saturated ground below the floor of the crater. Figure 6 shows the results of this simulation for a 100-km crater. The major differences seen in this run are due to the lack of pressure exerted by the crater lake, which results in water being drawn from the permeable right boundary rather than being resupplied from the lake. This causes the fluxes early in the system to be lower than before, and thus less heat is removed from the system. Unlike the previous model, there is no flow through the central peak and by 20,000 years most of the activity is concentrated in the center of the crater, which eventually develops into a single vigorous upwelling in the center of the crater by 200,000 years. This long-lived upwelling, coupled with the overall lower fluxes 10 of 19

11 Figure 5. Results of a numerical simulation of the hydrothermal system at a 180-km crater on early Mars. The center of the crater is on the left side of each figure. Surface permeability k 0 is 10 2 darcies. Solid arrows and dotted arrows indicate the water and steam fluxes, respectively. The lack of arrows indicates that fluxes are less than 2 orders of magnitude smaller than the maximum flux. Solid lines are isotherms, labeled in degrees Celsius. The length of the arrows scales logarithmically with the flux magnitude, and the maximum value changes with each plot. (a) 4,000 years, max. water flux = kg s 1 m 2, max. steam flux = kg s 1 m 2 ; (b) 20,000 years, max. water flux = kg s 1 m 2, max. steam flux = kg s 1 m 2 ; (c) 200,000 years, max. water flux = kg s 1 m 2 ; (d) 2,000,000 years, max. water flux = kg s 1 m 2. seen in this run, results in a longer system lifetime (700,000 versus 430,000 years) for this simulation compared to a case with the crater lake present Heat Transport in the Absence of Fluid Flow The 30-, 100-, and 180-km Craters [46] To understand the effects of water on crater cooling it is helpful to compare and contrast the simulations described above to identical scenarios but without water. HYDROTHERM can be used to model crater cooling using purely conductive heat transport in the rock matrix by setting permeability and porosity to near zero. Figures 7 9 show cooling of the 30-, 100-, and 180-km craters, respectively, in the absence of water. In the 30-km crater model, the temperatures in the 1,000 and 10,000 time steps are significantly higher due to the lack of heat removal by water. However, in the 100,000 year time step of the wet model, temperatures are actually higher in the crater s central region due to the heat deposited by a long-lived upwelling of warm water. These vertical upwellings also result in the time required to return to geothermal gradient being longer for the wet model compared to the dry model. [47] In the 100-km and 180-km crater models, a similar trend can be observed. While the dry model is on the whole hotter, there are regions in the wet model, such as the peak ring, where long-lived hot water upwellings have significantly increased the temperature. Conversely, there are numerous downwellings of cold water that led to a localized temperature decrease. These deflections in the 11 of 19

12 Figure 6. Results of a numerical simulation of the hydrothermal system at a 100-km crater on early Mars, without a crater lake. The center of the crater is on the left side of each figure. Surface permeability k 0 is 10 2 darcies. Solid arrows and dotted arrows indicate the water and steam fluxes, respectively. The lack of arrows indicates that fluxes are less than 2 orders of magnitude smaller than the maximum flux. Solid lines are isotherms, labeled in degrees Celsius. The length of the arrows scales logarithmically with the flux magnitude, and the maximum value changes with each plot. (a) 500 years, max. water flux = kg s 1 m 2, max. steam flux = kg s 1 m 2 ; (b) 4,000 years, max. water flux = kg s 1 m 2, max. steam flux = kg s 1 m 2 ; (c) 20,000 years, max. water flux = kg s 1 m 2, max. steam flux = kg s 1 m 2 ; (d) 200,000 years, max. water flux = kg s 1 m 2, max. steam flux = kg s 1 m 2. temperature contours are entirely absent in the dry model due to a lack of flowing water Hellas Basin (2,000 km) [48] The thermal evolution of a Hellas-sized impact basin is shown in Figure 10. While the vertical resolution of the model (30 km) for this immense structure is too coarse to show hydrothermal activity in the upper 10 km of the surface, this activity is not expected to have a significant effect on the cooling of the system of this magnitude. Thus only cooling by conduction is modeled. However, most of the hot spot is located in the mantle, but neither mantle convection nor convection in the melt produced by the impact is included in this model, which should be treated as a rough approximation. The model results indicate that the lifetime of the hydrothermal system produced by an impact of this magnitude would be on the order of 10 Myr, which is the time it takes for the upper 1 km of the surface to cool below 90 C (see next section for an explanation) Effects of Permeability on System Lifetime [49] Permeability is arguably the single most important parameter affecting the nature and duration of an impactinduced hydrothermal system. In addition to the main simulation set with a surface permeability of 10 2 darcies, values of 10 3 and 10 1 darcies were also tested, corresponding to the average permeability of the Earth s crust and midrange permeability of crystalline rocks, respectively. Qualitatively, observed fluxes, vertical extent of convection 12 of 19